Bridging Data and Art: Investigating Data-Art Connections in a Data-Art Inquiry Program

To cultivate students’ data literacy, data science learning often involves data practices. Lee et al. (2022) proposed a framework depicting six key dimensions of data practices: frame problem, consider and gather data, process data, explore and visualize data, consider models, and communicate and propose action. Rooted in the idea that data science education should commence with problem-solving to enhance a system (Wild & Pfannkuch, 1999), the initial data practice involves identifying problems that can be addressed with data. Once a problem is identified, the next step is to consider collecting data to solve it. In this process, Hardy et al. (2020) emphasized the importance of students actively producing data through the design and revision of data instruments, understanding the material context of data production, and utilizing instruments with awareness of their affordances and constraints. Processing data is another vital practice. For data processing, Erickson et al. (2019) proposed six “data moves” for data processing—filtering, grouping, summarizing, calculating, merging/joining, and making hierarchy—essential steps for data analysis that were often overlooked in data science curricula. The next practice is exploring and visualizing data. As outlined by Lee et al. (2022), this practice involves identifying relationships, patterns, and trends and using visualization to effectively communicate results. Following this step, data modeling transforms real-world questions into relevant data that is collected and analyzed to guide inferences about those questions (Pfannkuch et al., 2018). The final practice is to communicate and propose action, encompassing interpreting results and framing evidence-based claims. Besides, data-based storytelling is also a key component of communicating data. Echoing Pfannkuch’s (2011) notion of data context, which describes the situation-related data, the involvement of subject knowledge, and the knowledge that constructs data, data-based storytelling empowers students’ conceptual understanding and communication of data (Pfannkuch et al., 2010). More importantly, it enables students to tie data with representations that allow them to produce new structural information or contextual details related to the given situation (Wilkerson & Laina, 2018). Eventually, this process helps students to reason and make inferences with data (Rubin, 2020).

Data and art practices have shared qualities that render the integration of data practices and arts feasible. For data practices, Lee et al. (2021) envision three overlapping layers that make data practices transdisciplinary: personal, cultural, and sociopolitical. The personal layer includes students’ data experiences, interests, prior knowledge, and other personal aspects that may influence data practices (e.g., Lee & Dubovi, 2020). The second cultural layer involves the sociotechnical tools, artifacts, and cultural practices that guide students to collect, analyze, interpret, and communicate with data (e.g., Lehrer & Schauble, 2000). The third sociopolitical layer explores how student’s data science learning is shaped by power dynamics (e.g., Van Wart et al., 2020). These three layers can provide a framework for integrating data and arts. Art can serve as a medium for cultural and social transformation and promote transdisciplinary exploration and discovery (Guyotte et al., 2015), which aligns with the transdisciplinarity framed by the three layers of data practices. The personal, cultural, and sociopolitical layers are addressed in arts-based learning. In the personal layer, art production provides students with opportunities to create a meaningful personal space (Sakatani & Pistolesi, 2009); in the cultural layer, art bolsters students’ cultural participation (Nagel et al., 1997); and most importantly, the art making process fosters students’ sociopolitical consciousness (Ngo et al., 2017). The mapping in the three layers bridges the transdisciplinary nature between data practices and art production, consolidating the theoretical foundations of data-art integration.

In a data-art integration program, visualizing data plays an essential role in terms of learning outcomes, problem-grounding process, and idea presentation. First, regarding the final outcome, arts-based learning often involves the development of a performative outcome that manifests the processual and procedural importance of the learning journey (Marshall, 2014). Similarly, among those data practices, data visualization is a form of performative outcome that necessitates students’ previous input of data collection and analysis. Second, art inquiry allows students to make abstract ideas concrete (Colucci-Gray et al., 2017), and data practices also empower students to identify problems and collect and analyze data to ground the problem through data visualization. Lastly, data analysis and modeling require students to develop a deep understanding of their topics, such as establishing connections between different variables, and arts-based learning creates students’ deep knowledge by enabling them to know the central and crucial ideas of the topic and create connections between those ideas and artistic presentations (Robinson, 2013).

Given that arts can support learning and promote creativity in various STEM areas (Ge et al., 2015; Sousa & Pilecki, 2018) and the convergence of data and art practices, data-art inquiry emerges as a promising method in the field of data science education. Data-art inquiry aims to enhance students’ data literacy by engaging them in the processes of question generation, data collection and analysis, and using artistic methods to visualize and communicate data (Matuk et al., 2022). This approach enables students to communicate their understanding of data and critical issues artistically, ultimately enhancing their critical data literacy (Woods et al., 2024).

Several data-art inquiry programs were documented in the literature, shedding light on innovative approaches to enriching data science education. Matuk et al. (2022) introduced four data-art inquiry programs with middle schoolers. The first program was an 18-week dance program in which students used dance as a medium to communicate existing data on various topics. However, challenges arose as students occasionally veered toward art-based rather than data-based choreographic choices and consequently omitted the sociopolitical nature of their data-art inquiry. The second program spanned 3 weeks, during which students took photos of neighborhoods and used local public data to explore the elements contributing to a healthy neighborhood. This program had the potential to use data-art inquiry to address personal, cultural, and sociopolitical issues. However, in this program, students faced the challenge of synthesizing ideas across data, photos, and writing. In the third program, a 6-week art program, students employed self-generated survey data to create comics illustrating their friendship experiences. This program showed the integration of data and art at a personal level, whereas there was also tension in this program, with students encountering conflicts between art choices and data claims. The fourth program was a 1-week program in which students used public data about teenagers’ time use to make a collage to present the relationship between personal time use and wellbeing. Similar to the third program, this program also focused on exploration at a personal layer. Nonetheless, this program also faced challenges—students not significantly engaging in data discussions or resonating with datasets. Furthermore, Bhargava and D’Ignazio (2017) conducted a research project focusing on undergraduate and graduate students in a week-long program, wherein participants created data sculptures representing existing public data. Then, they tested out this approach in ten elementary schools to enable students to tell data stories with their data sculptures. While this project emphasized engagement with data and the potency of physical data visualizations for storytelling, it provided limited examples and details about how students integrated sculpting arts into their data narratives, leaving this aspect underexplored. Another project by Bhargava et al. (2016) included a 2-day “data mural” initiative in Brazil, where 20 students participated in a workshop generating qualitative and quantitative data about their school relationships and then painted murals in their neighborhoods based on the data they had gathered. This “data mural” project improved students’ data literacy, comfort with data, and interest in using data. However, as this study was preliminary and focused more on the implementation of the data mural program, more research is needed to explore how the artistic data learning process contributed to enhancing data literacy. These two projects (Bhargava & D’Ignazio, 2017; Bhargava et al., 2016) emphasized the implementation of data-art inquiry and students’ development of data abilities, but the transdisciplinary exploration of data-art inquiry was not explicitly presented. Although these studies successfully implemented data-art inquiry programs, there remains untapped potential to explore how students effectively combine data and art to address personal, cultural, and sociopolitical issues.

Aiming to cultivate data literacy among middle and high school students, we designed a 13-week data-art inquiry program called MVP (Mathematizing, Visualizing, and Power). The program encouraged students to identify community issues, collect/identify relevant data, and present these issues and data through artistic data visualizations. Reflecting the iterative nature of design-based research, the MVP program consisted of four cycles from Spring 2023 to Summer 2024, though this study focused solely on the first cycle in Spring 2023.

The MVP program, aimed at implementing a transdisciplinary data science education afterschool program, was developed by researchers from the University of Tennessee, educators from Knox County Schools, and community partners from the Knoxville area. The program included 12 learning sessions and three community learning events (see Fig. 1). Each weekly learning session was approximately 1.5 h in duration. Each session followed a similar format that included a warm-up activity introducing an art technique, a small-group exploration task, and a reflection. In these sessions, an instructional team led the activities and discussions while working with the student participants, referred to as “data artists.” Additionally, the program featured community learning events (CLEs) that took place at different locations in the city. CLEs were open to the public, and the data artists’ families and friends were invited to attend these hour-long events. Food was provided and data artists shared activities they had experienced as part of the learning sessions, collected data, demonstrated works in progress, or exhibited and discussed their final data visualizations. In Cycle 1, all learning sessions and two CLEs were held at a local after-school club, while the final CLE, designed as a grand finale, took place at an art center within a local community college.

Full size image

The project was publicized through university social media outlets for the department and the STEM education center with which the project was affiliated. In addition, the after-school club publicized CLEs by sending a flyer to the parents of all club members through the Remind app. In addition, two project team members attended two orientations hosted by the club at the beginning of the semester, which included representatives from a number of programs. From a get-to-know you survey shared with participants at the beginning of the learning sessions, it seemed that participants joined the project because they had an interest in art, STEM subjects, or both. Also, the stipend associated with project participation was meant to support youth motivation and retention. Most students’ prior art experiences included school art classes and/or an art program at the after-school club. Experience with data were mostly school-based.

The instructional team for the MVP program consisted of three primary instructors: one mathematics, one English, and one art instructor. The mathematics and English educators were recruited from a local public university’s teacher education program based on their interest and experience working with K-12 students. The art instructor was a local high school art teacher recommended by a member of the project team in consultation with the local school district’s art education leadership. In addition, there was a teaching assistant who was a preservice mathematics teacher from the local university. During the first two segments of the learning sessions, instructors delivered lectures, facilitated individual and group tasks, and orchestrated small-group and whole-group discussions; the facilitators then transitioned to conferencing and working with data artists individually and in small groups as data artists worked on their final individual/small group data visualizations.

Twenty-seven data artists participated in the MVP program. Since research participation was voluntary, eight data artists consented to our research. We invited all eight data artists for our post-program interview and six participated. Table 1 shows their demographics. Data artists were paired to be interviewed each time. Each interview was conducted online via Zoom and lasted approximately 1 h. All the interviews were completed within 2 weeks after the MVP program ended in May 2023. We asked eight primary questions (see Appendix) and some follow-up questions during each semi-structured interview.

ENA is the primary data analysis method in this study. ENA is an appropriate tool for investigating the co-occurrence and interconnections between qualitative codes. ENA has several advantages: it shows visual representations of how concepts are connected, combines qualitative and quantitative analysis, captures temporal changes, provides comparative analysis and contextual insights, and is suitable for interdisciplinary analysis; however, ENA requires high-quality qualitative coding, faces technical challenges while analyzing very large datasets, and risks interpretability issues when the network is of excessive complexity (Shaffer, 2017, 2018; Shaffer et al., 2016).

For our data analysis, our research question investigates the connections between data and art practices in a data-art inquiry program. ENA can illustrate the co-occurrence and interconnections between qualitative codes related to these two concepts. To get insights from ENA, we first needed to conduct qualitative coding of data and art and then built ENA models to explore the connections between those codes. Due to the exploratory nature of our research, we used an inductive coding approach (Saldaña, 2009) to develop a coding scheme. We first preliminarily coded the Zoom-generated transcripts in a document form. Then, reflecting on Lee et al.’s (2022) six data practice dimensions, we categorized those preliminary codes into eight themes and created a codebook (Table 2) to further explore transdisciplinary learning in the MVP program.

With the codebook, we used a web tool called Taguette to conduct qualitative coding. Taguette allows users to upload their codebook and tag codes to selected texts that are named highlights (Rampin & Rampin, 2021). After coding all the interviewees’ turns of talk, we exported highlights from Taguette in a CSV spreadsheet. We then used the Python pandas package to format coded results into a binary coding format (Fig. 2) to make it fit for ENA. On this formatted spreadsheet, we had four types of metadata—speaker, timestamp, turns of talk, and date—and for each row on this spreadsheet, 1/0 under each code column means that row has/does not have that code.

Full size image

Then, we imported the processed spreadsheet to the ENA web tool (https://www.epistemicnetwork.org/). To construct an ENA model, there were four components—unit, conversation, moving stanza window, and codes (see Table 3). In our ENA models, we defined the unit of analysis as a speaker, which consisted of all the lines associated with that speaker. The conversation is the local structure for the coded qualitative data, which represents a set of interrelated lines (Shaffer, 2017, 2018). In this study, since we conducted three interviews with three groups of data artists on three different dates and their turns of talk were related to the data visualization they created and their MVP program experience, we use the date to define the conversation of our ENA models. Within a conversation, ENA uses a moving stanza window to construct a network model for each line in the data, showing how codes in the current line are connected to codes that occurred previously (Ruis et al., 2019) and the size of the moving stanza window should best capture the recent temporal context in the recorded discourse (Shaffer, 2018). In our case, since our interview was a group interview, including one interviewer and two interviewees, the typical structure was the interviewer asking a question and the two interviewees responding to that question. Thus, the best moving stanza window size is three lines (one line plus two previous lines).

The codes in our ENA models were at two levels and varied depending on the level. One was at the theme level, which examined the connections between different themes, so the ENA models had eight themes—–Data collection, Data topic, Data pattern, Impact, Data interpretation, Data processing, Art technique, and Art design—as codes. The other was at the code level, which examined the connections of those codes as shown in the codebook under two specific themes. The themes were aggregated from codes.

Table 4 shows the data visualizations the interviewed data artists created, the title of those data visualizations, and their artist statements for the presentation in the final community learning event. For the first pair of data artists, Karen created a data visualization of mixed materials to show the data collected from her classmates and friends about their feelings about schooling; Suki designed a survey for people in the afterschool club where MVP took place to know more about people’s music preferences and presented her data visualization through an Instagram page format. For the second pair, Nina prepared questions for kids in the afterschool club to learn their parents’ strictness and their relationship and presented her analysis through canvas painting; Waj did not present a final data visualization. For the third pair, Bob and Jamos asked people about their favorite soccer players and whether soccer should be provided in high school curricula, and they painted a soccer ball and bounced and rolled it to present their data.

The theme-level ENA networks explore the connections between data practice nodes and art production nodes, providing an overview of each individual’s data-art connections. In each ENA network, each plotted point is a unit, the nodes are codes, and the edges reflect the relative frequency of co-occurrence between codes (edge weights are not real numbers but relative indicators of the strongness of different connections). The size of nodes maps the relative frequency of the codes (Shaffer, 2018). The unit of analysis of this study is each individual. We comprehensively examined each data artist’s theme-level network, investigating the edge weights between data practice codes and art production codes. This theme-level analysis revealed the distinctive approaches employed by each data artist in connecting data practices and artistic elements.

For Karen (Fig. 3a) and Suki (Fig. 3b), compared to other data practice codes, “Data topic” has stronger connections with codes “Art design” and “Art technique.” In Karen’s graph, the relative edge weights are both 0.49 (“Data topic-Art design” and “Data topic-Art technique”), and in Suki’s graph, the relative edge weights are 0.55 (“Data topic-Art technique”) and 0.47 (“Data topic-Art design”). These connections indicate that Karen and Suki’s art design and technique use are closely related to their data topics. Besides, in both Karen and Suki’s networks, the node size of “Data collection” is the second largest, indicating both of them stressed their data collection process in the interviews.

Full size image

For Nina (Fig. 4a), there are two strong connections—“Data collection-Art design” and “Data interpretation-Art design,” both with a relative edge weight of 0.6. These connections show that Nina’s art design for her data visualization was based on her data collection experience and her interpretation of her data visualization. As for Waj (Fig. 4b), the connections between data practice codes and art production codes were not strong overall, possibly because she talked extensively about data collection and the patterns she identified in the data, as shown from the node sizes of codes data collection and data pattern as well as the strong connection between these two codes.

Full size image

As for the soccer group, data collection has noticeable connections with art design and art technique. In Bob’s network (Fig. 5a), the relative edge weights are 0.73 for “Data collection-Art design” and 0.88 for “Data collection-Art technique,” which are more than any other data-art connections. Similarly, in Jamos’s network (Fig. 5b), the weight between codes “Data collection” and “Art design” is 0.8. These connections indicate their art design and technique use are closely associated with their data collection.

Full size image

Based on what we found from each data artist’s theme-level network, we examined these connections in the code-level networks to signify the connections between codes under those mutually related themes. For Karen and Suki, we examined “Data topic-Art design” and “Data topic-Art technique.” For Nina, we examined “Data interpretation and Art design.” For Jamos and Bob, we examined “Data collection-Art design” and “Data collection-Art technique.”

Data Topics and Art Design. Figure 6 is Karen’s network for the codes under themes “Data topics” and “Art design.” The edges that are pointed by the black arrows are the most noticeable connections, which are “Format-Personal interests” and “Format-Personal experiences,” both with a relative weight of 0.33. These two connections indicate that Karen’s personal interests and experiences inspired her art design for the final data visualization. During the interview, Karen talked about the topic of her project: young people’s feelings about middle and high school. She wanted her project to show a shared experience but also something individual, so she related the topic to school. In her data visualization (see Fig. 7), she used a pencil to represent each person and made it a “sun” or “clock” shape. The pencils connecting in the middle represented people were connected because they had similar experiences of schooling, but the pencils pointing out in different directions indicated people had their own experiences and feelings about schooling. In this case, her data topic determined the art design of her data visualization.

Full size image

Full size image

Data Topic and Art Technique. Figure 8 is Suki’s network for the codes under the themes “Data topic” and “Art technique.” The edge that is pointed by the black arrow between codes “Color coding” and “Personal interests” outweighed any other edges with a relative weight of 0.83. This connection indicates Suki’s personal interest determines her strategy of color coding while creating her data visualization, which is also shown in Excerpt 1.

Full size image

Excerpt 1: Suki: “Personally I’m a really big music person. I love music. I love learning about music, listening. I play guitar…For each genre I found most popular, I assigned a color so there was orange for rap, purple for R&B, and pink for pop… And I painted each box...There were like three oranges, one pink, one that I mixed the pink and the purple together because it was a rap and pop, R&B and pop…”

In this excerpt, Suki showed her strong passion for music and learning more about music, and she was familiar with musicians and genres from her survey respondents. While doing color coding, she knew that there might be a mixed genre, so she used a unique color-coding technique to mix paints for a new color to show the mixed genre (see Fig. 9). Suki’s data topic was chosen based on her strong interest in music, and her unique color-coding technique was built on her music interest and knowledge.

Full size image

Data Interpretation and Art Design. Figure 10 is one of Nina’s code-level networks showing the connections of codes between themes “Data interpretation” and “Art design.” The edge that is pointed by the black arrow is the strongest connection between codes “Meaning” and “Symbolic elements,” with a relative weight of 0.62. The meaning of her dataset inspired the art design of her data visualization. Nina’s data visualization was about the parent–child relationship. While creating this data visualization (see Fig. 11), she first identified that there were three elements in her dataset to visualize—parents and children, the strictness of parents, and the relationship between parents and children. Then, she designed symbols to represent these elements in a meaningful way: she distinguished parents and children in the data visualization through the thickness of the “i” symbol, defined the strictness of parents with the difference of the heights of those symbols, and distance symbols in a painted circle to indicate the relationship between parents and children. Her interpretation of the meaning in her data guided her symbol design, connecting her dataset and artwork.

Full size image

Full size image

Data Collection and Art Design. Figure 12 shows Jamos’s network of codes under the themes “Data collection” and “Art design.” The two highlighted connections that are pointed by the black arrows are “Results-Materials” and “Instruments-Abstractness,” both with a relative weight of 0.58.

Full size image

Excerpt 2: Jamos: “…Mbappé had the least. He had the least of all the answers so he went last…We used a lot of paint for him compared to some of the other ones where we didn’t need to use that level of paint.”

The connection “Results-Materials” means the results from Jamos’s data collection inspired his material use in the final data visualization. In Excerpt 2, he said Mbappé (one of the soccer players from the question who is your favorite soccer player) got the least vote of all answers, so they decided to use more paint for it intentionally to highlight he was another one besides two more popular players. The infrequent responses in the collected data can inspire data artists to think about how to make their design include all responses.

The other connection, “Instruments-Abstractness,” was about the artistic style of their data visualization. During the interview, Jamos said their data visualization (see Fig. 13) was more like abstract art and not from a survey. This statement indicates he had a predefined format of data visualization that was more associated with survey data, but their work deliberately countered this format. This data-art connection shows that data artists intentionally elevated the abstractness of their artwork to avoid their data visualization associated with survey data, indicating that data artists connected the instrument of data collection with a form of data visualization associated with it.

Full size image

Data Collection and Art Technique. Figure 14 is Bob’s network of codes under the themes “Data collection” and “Art technique.” The code “Results” has two strong connections (pointed by the black arrows) with codes “Bounce and roll” and “Color coding,” both with a relative weight of 0.76. These connections imply Bob’s primary art techniques are bouncing and rolling the soccer to color code the raw results from their data collection, as exemplified in Excerpt 3.

Full size image

Excerpt 3: Bob: “…We had red represent Ronaldo, light blue represent Messi, dark blue represent Mbappé, orange represent Yes, and then green represent No…So we put the paint on top of the ball…We bounced on the paper. And what we did was, for the last [answer], we rolled it instead of bouncing it…”

Bob collaborated with two other data artists on a data visualization about people’s favorite soccer players and whether soccer classes should be provided in high school. In Excerpt 3, he talked about their use of colors for different favorite players for the first question and “yes” or “no” answers for the second question. They put paint on the soccer ball and bounced it with different colors to represent the votes of three players and the counts of “yes” or “no” answers and rolled for the last count of each category (see Fig. 13). This excerpt reveals that he associated the use of color and the technique of bouncing and rolling with counts of different categories from the raw data.

This examination of how data artists connect data practices and art production reveals three distinct foci: topic-inspired data-art connections, meaningful-symbol-based data-art connections, and innovative art technique-based data-art connections. Karen and Suki drew inspiration for their artistic data visualization from their closely relevant data topics that were rooted in their previous experiences and personal interests. This connection aligns with Lee et al.’s (2021) personal layer, as their data topics were based on their personal experiences and interests, driving them to reflect on their experiences through art and showcase their passions. Nina, on the other hand, focused more on interpreting the results from her data, using symbols as culturally meaningful expressions to connect data and art. Her data visualization showcased a potential community issue, the child-parent relationship, via a symbolized way. Symbols are culturally meaningful as a system to communicate and develop people’s knowledge and attitude toward life (Geertz, 1972). Nina’s use of meaningful symbols portrays her interpretation of the child-parental relationship issues in her community and potentially raises others’ concern about this issue, which successfully reflects Lee et al.’s (2021) cultural layer. In addition, Bob and Jamos employed an innovative approach—bouncing and rolling a painted soccer ball—to create their data visualizations based on the raw data. This unique technique, aiming to ensure the inclusion of less famous players, aligns with Lee et al.’s (2021) cultural layer, as they sought to use this distinctive art technique in a soccer-unpopular context to highlight the importance of all data points in their dataset. At the same time, the soccer data visualization was personally relevant to Bob and Jamos, since they played soccer together and were keen on trying different painting techniques with the soccer ball. Despite these differences, all data artists’ combinations of data and arts resonated at both the personal and cultural layers, making their data-art inquiry experience meaningful.

Among the identified data-related practices, data collection emerged as the most essential in the data-art inquiry program. Evident in the theme-level networks, Data collection is either the largest or the second largest node and is located close to the two art production nodes. This centrality indicates the resonance of data artists with raw data and its subsequent influence on their art production. The data collection process allowed them to experience how the data was created, utilize their previous knowledge to understand the results, and then develop corresponding art approaches to visualize their data. Besides, engaging in data collection thus became a focal point for data artists, linking this practice directly to the other essential aspect—art production. Furthermore, data collection involved data artists’ interaction with peers in their communities and their final artwork was also for their community members. Their data collection process involved their future artwork audience, and as a result, their artwork became more communicative and relevant to their community members.

Lastly, the role of art process in a data-art inquiry program is irreplaceable, from both art design and art technique perspectives. In a data-art inquiry program, instead of superficial engagement with art, the data-driven art production promotes “the dispositions and meaning-making practices inherent to art (Bertling et al., 2021, p. 44).” From an art design viewpoint, art processes provided a unique stage for data artists to create their “personal space” (Sakatani & Pistolesi, 2009), transcending the limitations of traditional data visualization types, such as bar chart and line graph. For example, Karen’s innovative sun-shaped visualization reflects her personal perspective on schooling, demonstrating the power of art for constructing individual narratives. From an art technique perspective, color-coding emerges as a common approach, fostering efficient information acquisition (Keller & Grimm, 2005). In addition, specific art techniques, such as bouncing and rolling a painted soccer ball or drawing on pencils, not only reflect data artists’ understanding of their dataset but also contribute to the creation of unique and meaningful data visualizations at personal and cultural levels.

In summary, this study illuminates the nuanced connections between data practices and art production, underscores the pivotal role of data collection, and emphasizes the contribution of art production in expressing individual narratives within a data-art inquiry program.

This study offers three practical contributions to inspire the future design of data-art inquiry programs. First, we recommend providing diverse pathways for data artists to integrate data and art. Considering the varied backgrounds and prior knowledge of students, allowing them to explore their own data topics, conduct independent analyses, and visualize datasets using art design and techniques of their choosing can make the program personally meaningful and culturally relevant. Second, we suggest emphasizing the role of data collection. Compared to the previous data-art inquiry programs that often relied on preexisting data sources, the MVP program focused on students collecting their own data. By using self-designed instruments to gather data, students enhance their understanding of their dataset and create a deeper connection between their collected data and their artistic modes of communication. Lastly, we encourage tailoring the duration of data-art inquiry programs to accommodate participants with varying levels of data science and art backgrounds. The process of collecting, analyzing, and visualizing data can be overwhelming for those with limited knowledge. The MVP program allows participants to gradually familiarize themselves with data and art skills, building data-art connections progressively and providing a more manageable and structured learning experience.

Additionally, we presented a model for implementing a long-term, afterschool, data-art inquiry program that offers students extensive experience in data practices, community topic exploration, and art production. Compared to previous data-art inquiry programs, the MVP program integrates art and data work, rather than using art merely as a tool for presenting data. Moreover, the MVP program emphasizes the relevance of the data to the students and their communities by cultivating the meaningfulness of their projects at a personal and communal level, which aligns with Lee et al.’s (2021) humanistic data science education.

There are also limitations in this study. Firstly, this study primarily explores the connections between data practices and arts, but, from the theme-level network, some connections between data practice themes are also strong, especially between themes Data collection and Data pattern, so are the connection between two art production themes. We did not explore those connections here, but it can be interesting to investigate those intra-theme connections. Second, our data artists did not talk about the sociopolitical impacts of their data visualization considerably. Since the first MVP cycle focused on the meaningfulness and relevance of data and art for data artists, in the following iteration, we planned to encourage students to think about critical data topics that address their community issues. Third, we only interviewed six data artists in three paired interviews due to consent status and scheduling issues. Our findings are based on these three interviews, but a larger sample and the inclusion of their conversations during the learning sessions and community learning events might have yielded different findings. Fourth, in this study, we only explored the strong connections of each individual. However, defining a strong connection with the thickness of edges and numeric weight values on the ENA networks risks omitting other key connections that are critical yet less mentioned in the coded discourse data. For example, a student may actually spend a lot of time processing and formatting the data but barely mention that process during the interview because they think data processing is tedious and not worth mentioning. Thus, in the future study, more data sources should be added to the analysis to improve the reliability of findings from the ENA networks.